In recent years, from the security viewpoint, it is becoming important to use a surveillance camera to perform surveillance in order to discover a suspicious person and/or a suspicious substance, for example. In order to reduce the load on a watchman who watches the video photographed by a surveillance camera, various systems have been disclosed in which a computer may be used to image-process the video by the camera to discover a suspicious person and/or a suspicious substance.
For example, a technology has been disclosed which extracts an object by comparing a photographed image and a pre-photographed past background image as discussed in Japanese Laid-open Patent Publication No. 11-32325, and another technology has been disclosed which irradiates pattern light, compares an image with an object and an image without it and extracts the object on the basis of the difference in pattern as discussed in Japanese Laid-open Patent Publication No. 2003-269915.
However, both of the conventional technologies have a problem that an object may not be extracted with high accuracy. FIG. 17 is a diagram for describing an example that the color of the object to be extracted and the background color are similar. FIG. 17 is a plan view illustrating objects 83 to 85 and a background 82. The object 83 includes parts 83a and 83b, the object 84 includes parts 84a and 84b, and the object 85 includes parts 85a and 85b. The objects 83, 84 and 85 have different colors. More specifically, according to Patent Document 1, as illustrated in FIG. 17, when the color of the entire object is much similar to the background color, the object may not be extracted. For example, in FIG. 17, it is difficult to extract the object 84 in the lower part of the diagram.
According to Patent Document 2, due to the effect of the shade of the pattern light, the object may not be extracted clearly. Furthermore, according to Patent Document 2, because the shaded object may not be extracted, the extraction of plural objects is not supported.